Every business in this study was found the way a customer finds it: on Google Maps, searching for a plumber, a dentist, an accountant or a café in a named suburb. The listing was kept whether or not it had a website, and every website was put through the same measurements. This is what came back.
businesses listed on Google Maps in the region has no usable website. 11% have nothing at all and 2% point their listing at a Facebook or Instagram page.
- 57median mobile performance score, out of 100n = 736
- 46%have a tap-to-call button in the first phone screenn = 735
- 34%carry LocalBusiness structured datan = 742
- 39%show their opening hoursn = 742
- 15%have enough traffic for Chrome to report real-world speedn = 737
- 59%run any analytics at alln = 742
Who has a website
Across 1,030 listings, 86% point at a real website. The rest split between nothing at all (11%), a social page standing in for a site (2%) and a directory profile (fewer than 1%).
The gap is not evenly spread. 5% of health listings and 7% of professional firms lack a site. In hospitality and personal services it is 24%, and the single worst category is hair salons at 33%. Among trades, painters (23%) and landscapers (25%) are the most likely to be running on a phone number and a Facebook page.
Table view
| Category | Listings | No usable website |
|---|---|---|
| hair salon | 45 | 33% |
| cafe | 74 | 28% |
| landscaper | 44 | 25% |
| painter | 47 | 23% |
| cleaner | 44 | 21% |
| mechanic | 42 | 19% |
| pest control | 43 | 19% |
| roofing | 41 | 17% |
| electrician | 52 | 15% |
| chiropractor | 36 | 14% |
| plumber | 48 | 13% |
| builder | 47 | 11% |
| accountant | 48 | 10% |
| solicitor | 42 | 10% |
| gym | 43 | 7% |
| air conditioning | 41 | 5% |
| real estate agent | 43 | 5% |
| physiotherapist | 54 | 4% |
| veterinarian | 41 | 2% |
| mortgage broker | 45 | 2% |
| ndis provider | 46 | 2% |
| dentist | 48 | 2% |
Review count tracks it closely. Of listings with no Google reviews, 37% have no usable website. Past 200 reviews, 4% do. That is an association, not a cause: businesses that invest in one tend to invest in the other.
What they are built on
WordPress powers 49% of the audited sites, against 36% of all sites on the wider web. Hosted builders (Squarespace, Duda, Wix, GoDaddy, Weebly, Webflow) account for 30%, and 18% are custom or hand-coded, most of them small static sites with no platform fingerprint at all. Among WordPress sites that use a page builder, Elementor dominates at 49% of them.
25% of all sites publish their WordPress version in the page source. 63% still load jQuery. 31% serve any image in a modern format such as WebP or AVIF. Why the platform decision sets the ceiling for everything downstream is the subject of WordPress vs Astro.
Table view
| Platform | Share |
|---|---|
| WordPress | 49% |
| Custom (hand-coded) | 12% |
| Squarespace | 9% |
| Wix | 9% |
| Duda | 7% |
| GoDaddy Website Builder | 3% |
| Custom (Vite) | 2% |
| Weebly | 1% |
| Webflow | 1% |
| Custom (Next.js) | 1% |
11.2% of sites credit a builder in the footer, 59 different agencies and freelancers in all. The three most-credited account for 20.5% of the credited sites. Nobody dominates.
Speed on a phone
Every site was run through Google's Lighthouse on a simulated mid-range phone over a throttled 4G connection, the same free test anyone can run in PageSpeed Insights. The median performance score is 57. 34% score under 50, which Lighthouse labels poor, and 4% reach 90.
The weight tells you why. The median page transfers 2.7 MB, against a web-wide median of 2.3 MB. Images are the largest part at 704 KB, then JavaScript at 959 KB. Third-party code (fonts, tracking, chat widgets, embedded maps and reviews) is 1.4 MB of a typical page. 21% of sites are over 5 MB.
Under those lab conditions the median largest contentful paint, the moment the main content is on screen, is 8.9 s. Google's target is 2.5 seconds; 4% of sites meet it in the lab.
Table view
| Score band | Sites |
|---|---|
| 0–9 | 0 |
| 10–19 | 14 |
| 20–29 | 39 |
| 30–39 | 87 |
| 40–49 | 119 |
| 50–59 | 179 |
| 60–69 | 158 |
| 70–79 | 90 |
| 80–89 | 48 |
| 90–100 | 34 |
Only 15% of these sites get enough Chrome traffic for Google to publish real-world speed data at all. Of those that do, 63% pass Core Web Vitals and their median real-world LCP is 2.1 s, far better than the lab figure because most visitors are on good connections. The other 85% are flying blind on speed, and so is Google.
Can a customer act
Each site was loaded on a phone-sized screen and the first screen inspected. 75% of sites have a tap-to-call link somewhere, but only 46% have one in that first screen. 241 sites have the link and hide it below the fold. A visible call to action of any kind sits in the first screen on 63% of sites.
Beyond the phone: 53% have a working contact form on the home page, 39% show opening hours, 60% show a street address, 44% show a price anywhere on the site. 21% use an online booking tool and 10% run a chat widget. 23% surface their Google reviews on the site.
Trades do this best: 60% of trade sites put the number on the first screen. Hospitality and personal services, where a customer wants to book or find a menu, manage 18%, and 47% show hours. What a tradie's site has to do in that first screen is the whole argument of web design for tradies.
Table view
| Feature | Trades | Health | Professional | Hospitality and personal | Other |
|---|---|---|---|---|---|
| Tap-to-call, first screen | 60% | 50% | 34% | 18% | 52% |
| Contact form on home page | 58% | 46% | 52% | 43% | 59% |
| Opening hours | 32% | 59% | 21% | 47% | 42% |
| Street address | 40% | 81% | 77% | 62% | 57% |
| A price anywhere | 30% | 55% | 60% | 64% | 27% |
| Google reviews shown | 29% | 22% | 22% | 8% | 27% |
Does Google know where you are
Someone in Belmont does not search "plumber Newcastle". They search "plumber Belmont", and Google can only show them a business it can place there. Counting every Newcastle, Lake Macquarie and Maitland suburb named anywhere across the home page and up to thirty inner pages, the median site names 2. 9% name none at all, and 38% name one or none. A dedicated suburb or service-area page exists on 44% of sites.
62% put a place name in the page title, and 31% in the main heading. Health practices are best at it (80% of titles); hospitality worst (52%).
Structured data is the other half. 76% of sites carry some schema.org markup, mostly generated by a WordPress SEO plugin, but only 34% declare a LocalBusiness with an address and phone, the block that feeds the knowledge panel and AI answers. Web-wide, LocalBusiness appears on about 4% of pages, so the region is well ahead there. Where the site shows a phone number, it matches the Google listing 77% of the time. 16% embed a map. The practical version of all this is in local SEO for Newcastle businesses.
Table view
| Signal | Newcastle | Lake Macquarie | Maitland |
|---|---|---|---|
| Place name in title | 68% | 61% | 57% |
| LocalBusiness schema | 36% | 34% | 31% |
| Phone matches Google listing | 80% | 74% | 77% |
| Opening hours | 38% | 38% | 43% |
| Street address | 64% | 56% | 61% |
| Map embedded | 16% | 16% | 17% |
Ready for AI search
AI answer engines read websites differently from Google: they favour plain text, direct answers, and a machine-readable summary at /llms.txt. 26% of sites in the region have one, which sounds like adoption until you look at where they come from. 77% of them were generated by the platform: Duda writes one for every site it hosts, and Wix, Squarespace and the WordPress SEO plugins do the same. The hand-written ones are a rounding error.
The robots.txt file is where a business tells AI crawlers what they may take. 15% of sites say anything at all about any AI crawler; 6% block at least one. The other 85% are silent, which means every crawler is welcome and none is guided. Web-wide, 7.9% of sites block GPTBot and 10.3% block at least one AI crawler, so the region blocks less than the web and says less.
5% use FAQ structured data, and 28% have even one heading written as a question, the form an answer engine lifts most readily. What being cited by an AI answer actually takes is on the GEO Newcastle page.
Table view
| Crawler | Allowed | Blocked | Not mentioned | No robots.txt |
|---|---|---|---|---|
| GPTBot | 10% | 6% | 79% | 6% |
| ClaudeBot | 10% | 5% | 79% | 6% |
| PerplexityBot | 1% | 1% | 92% | 6% |
| Google-Extended | 10% | 5% | 80% | 6% |
| CCBot | 10% | 6% | 79% | 6% |
| Bytespider | 9% | 5% | 80% | 6% |
| Applebot-Extended | 9% | 5% | 80% | 6% |
| OAI-SearchBot | 1% | 1% | 92% | 6% |
Neglect
A website is a thing you own, and the study caught plenty of signs of ownership lapsing. Of the 396 sites that print a copyright year, 27% show 2024 or earlier. Of the 93 with a dated blog post, 56% have not published in a year; the median gap is 15 months. 30% of sites have a blog at all.
17 listings on Google Maps point at a domain that no longer resolves or has been parked for sale, and 39 in all could not be reached by either a plain request or a browser. 12% of sites use a free email address (gmail, bigpond, hotmail) as the business contact. 15% still carry the old Universal Analytics tag, which stopped collecting in 2023. Of the 188 WordPress sites that reveal their version, 20% are more than a year behind. This is the case for a support plan in one paragraph.
Accessibility and measurement
The median Lighthouse accessibility score is 89, which sounds fine until you read the individual audits. 66% of sites have text that fails the colour-contrast standard (web-wide, WebAIM finds it on 79% of the top million home pages). 56% have links with no discernible name, 15% have images without alt text, and the median site describes 69% of its images.
41% of sites run no analytics of any kind: no Google Analytics, no Tag Manager, no Meta pixel. 27% have GA4. 31% carry a Search Console verification tag, the only free window into what Google actually shows them for.
Newcastle, Lake Macquarie, Maitland
The three council areas are more alike than different. Most measures sit within a few points, inside the margin. The differences that do show: Maitland has the highest share of listings with no usable website (17% against 11% in Newcastle), Newcastle sites are the heaviest (median 2.9 MB) and the most likely to run analytics (67%).
| Measure | Newcastle | Lake Macquarie | Maitland |
|---|---|---|---|
| Listings with no usable website | 11% | 13% | 17% |
| Median mobile performance score | 55 | 58 | 57 |
| Median page weight | 2.9 MB | 2.5 MB | 2.7 MB |
| Tap-to-call in first screen | 47% | 47% | 45% |
| Opening hours shown | 38% | 38% | 43% |
| LocalBusiness schema | 36% | 34% | 31% |
| Place name in title | 68% | 61% | 57% |
| Median suburbs named | 2 | 2 | 2 |
| llms.txt present | 24% | 26% | 27% |
| Any analytics | 67% | 54% | 58% |
| Contrast failure | 66% | 64% | 71% |
| WordPress share | 49% | 46% | 52% |
What they are missing
Ranked by how many sites lack it. None of these needs a rebuild. Most are an afternoon's work on the site a business already has. The ones that turn a search into a phone call are the phone number in the first screen, the opening hours, and the suburbs the business actually serves.
- 85%lack itEnough traffic to register in Chrome's field dataMost local sites are too quiet for Google to measure their real-world speed.
- 85%lack itA stated position on AI crawlers in robots.txtSilence means every AI crawler may take the content, and none is told where to find the good parts.
- 74%lack itAn llms.txt fileThe plain-text summary AI crawlers look for. Most that exist were generated by the platform, not written.
- 66%lack itText that meets the contrast standardThe most common accessibility failure on the web, and here too.
- 66%lack itLocalBusiness structured dataThe machine-readable card that tells Google and AI engines what the business is, where it is and how to ring it.
- 61%lack itOpening hours on the siteThe second most common question after "where are you".
- 54%lack itA tap-to-call link in the first phone screenHalf of every call to a tradie starts on a phone. If the number is not on the first screen, the visitor scrolls, or leaves.
- 48%lack itA contact form on the home pageMeasured after JavaScript ran, so builder-injected forms count.
- 41%lack itAny analytics at allNo way to know whether the site does anything.
- 38%lack itMore than one suburb named anywhere on the siteGoogle cannot rank a business for a suburb the site never mentions.
- 38%lack itA location named in the page titleThe title is the first thing Google reads and the first line of the search result.
- 34%lack itA mobile performance score of 50 or betterBelow 50 Lighthouse calls it poor.
- 32%lack itAn email on the business's own domainA bigpond or gmail address on a business site.
How it was done
Frame. Google Places, searched for 23 categories across six localities in each of Newcastle, Lake Macquarie and Maitland on 7 September 2026, 414 searches in all. Fifteen listings per category per council area, chosen round-robin across the six localities by Google's own rank so no town centre dominates. Every listing was kept whether or not it had a website. A listing was assigned to a council area by the locality in its address.
Exclusions. 32 national brands and franchisors whose listing points at a corporate site, 2 government or education sites, 1 site built by Newcastle Web Co, and 34 sites that could not be fetched. Every exclusion is counted, none is silent.
Measurement. Each site's home page, robots.txt, sitemap, llms.txt and up to thirty inner pages were fetched by an identified crawler. Each home page was then loaded in headless Chrome on a phone-sized screen to measure what is in the first screen and what only exists after JavaScript runs. Each was run once through Google's PageSpeed Insights (Lighthouse 13.4.1) on the mobile setting; very low scores and errors were run a second time and the better kept. Real-world speed comes from the Chrome UX Report where Google publishes it.
Validation. A 30-site sample, six per group, was checked by hand against the stored pages and the phone screenshots. Before correction, 168 of 180 field checks agreed (93%). The four detectors that fell short were corrected and re-applied to every site: platform, opening hours, parked domains, and what counts as visible in the first screen. After correction, 177 of 180 agree (98%); the three left are judgement calls the rules cannot make.
Reporting. Percentages are of sites where the field could be measured and n is shown beside each chart. Nothing is reported for a segment under thirty sites. The whole-sample margin at 95% confidence is about ±3.6 points; council-area cuts about ±6; category cuts are directional. Lighthouse scores are single lab runs and carry roughly ±5 to 10 points of run-to-run noise. Associations are reported as associations. No business is named, and the data behind every chart is aggregate only.
Limits. The frame is what Google Maps returns, so a business with no listing is outside it, and the sample is the fifteen listings per category Google ranks first in each council area, not every business in the region. That skews the sample toward businesses whose listing already works, so the true share with no usable website across all local businesses is almost certainly higher than one in 7, not lower. Detection is by fingerprint on the pages fetched; a feature that exists only on an unfetched page is missed. Lighthouse ran from Google's servers, not from the Hunter. Benchmarks from the Web Almanac and WebAIM sample the whole web, weighted to large sites, and are context, not a control group.
Benchmarks cited. HTTP Archive Web Almanac 2024 (CMS, Performance, Page Weight and Structured Data chapters); WebAIM Million 2025; the Australian Bureau of Statistics, Characteristics of Australian Business 2024–25; HasData's AI Crawler Block Index and Chris Humphrey's robots.txt study of the top 10,000 sites. The aggregate data behind this page and the scripts that produced it are available on request, so anyone can repeat the study. Cite it as Newcastle Web Co, The State of Small Business Websites in the Hunter, 2026, 7 September 2026.
Newcastle Web Co, Cameron Park NSW 2285. Edition one, 7 September 2026. Edition two, September 2027, same URL.